The entry point of the project is the run.py file. It is responsible for parsing the command line arguments and calling the appropriate functions for task 1 and task 2.

The run.py contains a class called FaceRecognizer that does the following:

• Loads the dataset
• Loads the patch shapes
• Loads the model for localization (task 1)
• Loads the model for classification (task 2)
• (Optional) Saves the results in pickle files
• (Optional) Checks the results with the ground truth

Localization consists of two stages:

• performing a sliding window over the image and classifying each patch as face or non-face
• performing post-processing on the results of the previous stage in order to obtain the bounding boxes of the faces and the corresponding confidence scores

Classification consists of one stage:

• classifying each face bounding box from the previous stage as one of the 5 classes (Barney, Betty, Fred, Wilma, Unknown)

To run the project you need to have installed the following packages:

• numpy==1.24.3
• opencv-python==4.8.1.78
• torch==2.0.1
• torchvision==0.15.2
• pickle
• matplotlib==3.8.2
• tqdm==4.65.0

You can install these dependencies by running the following command:

pip install -r requirements.txt

To run run.py you need to provide the following arguments:

• -d or --dataset_path(required): the path to the dataset
• -p or --patch_shapes_path(required): the path to the patch shapes
• -ml or --model_localization_path(required): the path to the model for localization
• -mc or --model_classification_path(required): the path to the model for classification
• -c or --process_config_path(required): the path to the configuration file for the post-processing
• -s or --save_results(optional): whether to save the results or not
• -g or --ground_truth_path(optional): the path to the ground truth file
• -h or --help: show the help message

Note that in the same directory as run.py there are the following files:

• model_localization.pth: the model for localization (task 1)
• model_classification.pth: the model for classification (task 2)
• process_config.pkl: the configuration file for the post-processing
• patches.pkl: the patch shapes

This is the full command for running the project on task 1 and task 2:

python3 run.py -d ../validare/validare/ -p ./patches.pkl -ml ./model_localization.pth -mc ./model_classification.pth -c ./process_config.pkl -s ../evaluare/fisiere_solutie/ -g ../validare/

This is the full command for running the project on bonus:

python3 run.py -d ../validare/validare/ -m ./model_resnet.pth -s ../evaluare/fisiere_solutie/ -g ../validare/